Comparison of liquid chromatography-isotope ratio mass spectrometry (LC/IRMS) and gas chromatography-combustion-isotope ratio mass spectrometry (GC/C/IRMS) for the determination of collagen amino acid δ13C values for palaeodietary and palaeoecological reconstruction

Research Article

Received: 17 April 2011

Revised: 30 June 2011

Accepted: 1 July 2011

Published online in Wiley Online Library

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

(wileyonlinelibrary.com) DOI: 10.1002/rcm.5174

Comparison of liquid chromatography–isotope ratio mass

spectrometry (LC/IRMS) and gas chromatography–combustion–

isotope ratio mass spectrometry (GC/C/IRMS) for the determination

of collagen amino acid d¹³C values for palaeodietary and

palaeoecological reconstruction^†

*

Philip J. H. Dunn, Noah V. Honch and Richard P. Evershed

Organic Geochemistry Unit, Bristol Biogeochemistry Research Center, School of Chemistry, University of Bristol, Cantock’s

Close, Bristol BS8 1TS, UK

Results are presented of a comparison of the amino acid (AA) d¹³C values obtained by gas chromatography–

combustion–isotope ratio mass spectrometry (GC/C/IRMS) and liquid chromatography–isotope ratio mass spectro-

metry (LC/IRMS). Although the primary focus was the compound-speciﬁc stable carbon isotope analysis of bone

collagen AAs, because of its growing application for palaeodietary and palaeoecological reconstruction, the results

are relevant to any ﬁeld where AA d¹³C values are required. We compare LC/IRMS with the most up-to-date

GC/C/IRMS method using N-acetyl methyl ester (NACME) AA derivatives. This comparison involves the analysis

of standard AAs and hydrolysates of archaeological human bone collagen, which have been previously investigated

as N-triﬂuoroacetyl isopropyl esters (TFA/IP). It was observed that, although GC/C/IRMS analyses required less

sample, LC/IRMS permitted the analysis of a wider range of AAs, particularly those not amenable to GC analysis

(e.g. arginine). Accordingly, reconstructed bulk d¹³C values based on LC/IRMS-derived d¹³C values were closer to

the EA/IRMS-derived d¹³C values than those based on GC/C/IRMS values. The analytical errors for LC/IRMS AA

d¹³C values were lower than GC/C/IRMS determinations. Inconsistencies in the d¹³C values of the TFA/IP

derivatives compared with the NACME- and LC/IRMS-derived d¹³C values suggest inherent problems with the

use of TFA/IP derivatives, resulting from: (i) inefﬁcient sample combustion, and/or (ii) differences in the intra-

molecular distribution of d¹³C values between AAs, which are manifested by incomplete combustion. Close

similarities between the NACME AA d¹³C values and the LC/IRMS-derived d¹³C values suggest that the TFA/IP

John Wiley & Sons, Ltd.

Collagen is an important structural protein in many tissues

and is the major organic fraction of skeletal remains, which

has obvious implications for archaeology and palaeoecology

because of its widespread recovery during excavations.^[1]

The bulk stable isotope analysis (SIA) of archaeological and

sub-fossil skeletal collagen (d¹³C and d¹⁵N values) using an

elemental analyser coupled to isotope ratio mass spectrome-

try (EA/IRMS) has become an accepted method of

reconstructing ancient human and animal diets.^[2–6]Whilst

the technique can deliver general dietary information for

numerous individuals within a rapid timeframe, and can

distinguish between populations with signiﬁcantly different

dietary histories (e.g. terrestrial C₃consumers vs. terrestrial

C₄consumers in the absence of marine products), bulk SIA

is not particularly sensitive to the detection of minor

contributions of isotopically distinct dietary components. It

is challenging, for instance, to identify low contributions of

marine food, or C₄plants in a predominately terrestrial C₃

diet, due to uncertainties in the dietary end-member values

for ‘pure’ C₃, C₄and marine.^[7]Moreover, bulk SIA is

typically limited to two variables per sample (d¹³C and

d¹⁵N), thereby limiting the parameters available to

distinguish between diet groups.

In recent years, individual amino acid (AA) stable

isotope analysis has been employed to enhance

palaeodietary research, not least because an understanding

of the isotopic character of skeletal collagen AAs is

essential to describing the origins of bulk collagen d¹³C

values.^[8–14]AA stable isotope analysis also has the

potential to elucidate observed differences in the isotopic

composition of body proteins, to understand how food

components are used in the production of body tissues,

and to derive dietary information from diagenetically

modiﬁed proteins.^[10,11,15]Given the promise of AA stable

* Correspondence to: R. P. Evershed, Organic Chemistry Unit,

Bristol Biogeochemistry Research Center, School of Chem-

istry, University of Bristol, Cantock’s Close, Bristol BS8

1TS, UK.

E-mail: r.p.evershed@bristol.ac.uk

†

Presented at the Liquid Chromatography-Isotope Ratio

Mass Spectrometry Users Meeting, held 23–24 November

2010 at the University of Oxford, Oxford, UK.

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

P. J. H. Dunn, N. V. Honch and R. P. Evershed

isotope analysis, interpretive frameworks are being

developed to synthesise a growing body of compound-

speciﬁc isotope data.^[8,9,16,17]The gas chromatography–

combustion–isotope ratio mass spectrometry (GC/C/IRMS)

analysis of skeletal collagen AAs, for instance, has revealed a

proxy for the detection of marine protein consumption, where

differences in the d¹³C values between the AAs glycine (Gly)

and phenylalanine (Phe) are substantially greater for consu-

mers of marine foods than for those consuming primarily

terrestrial-based resources (Δ¹³C_{Glycine-Phenylalanine}).^[9]This

ﬁnding has stimulated the investigation of other proxies to

identify the same phenomenon.^[8,16]Despite these advances,

it is clear that the interpretive frameworks ultimately depend

on a ﬁrm understanding of the capabilities of the analytical

technologies that are used to produce compound-speciﬁc

data: an area in which surprisingly little comparative work

has been devoted, but which this paper aims to address.

The compound-speciﬁc stable isotope analysis of skeletal

collagen AAs was ﬁrst achieved by Abelson and Hoering,^[18]

who utilised liquid chromatographic techniques to isolate

small quantities of pure, individual AAs. These were

combusted off-line, and the resulting CO₂gas was isolated

and puriﬁed for isotope ratio mass spectrometry (IRMS)

analysis. Similar off-line methods were later used in other

studies, including by Hare and Estep,^[19]Hare et al.,^[20]Macko

et al.,^[21,22]and Tuross et al.^[23]These studies have shown that

body proteins typically have an intra-individual range of AA

d¹³C values of ca. 20%.

The advent of gas chromatography (GC) coupled to on-

line IRMS via a combustion interface (gas chromatogra-

phy/combustion/isotope ratio mass spectrometry, GC/

C/IRMS) was ﬁrst reported by Barrie et al.^[24]GC/C/IRMS

is capable of obtaining d¹³C values for very small amounts of

analyte injected on column, as shown by Engel et al.,^[25]who

used the technology to determine the d¹³C values of AAs

from the Murchison meteorite, where the use of off-line

methods was precluded due to the low concentrations of

AAs present; other studies of AAs have since followed.^[26,27]

Critically, whilst GC/C/IRMS is capable of providing both

d¹³C and d¹⁵N values, AAs must be derivatised to be

amenable to GC. This process adds a degree of complexity

to the measurement of carbon isotope ratios speciﬁcally, as

derivative carbon contributes to the measured d¹³C values.

Furthermore, a kinetic isotope effect (KIE) will be introduced

wherever an acetylation step is used to derivatise the

N-terminus (or hydroxyl moiety) of an AA.^[28]However,

endogenous AA d¹³C values can be calculated using

correction factors obtained from the GC/C/IRMS-

determined d¹³C values of derivatised reference AAs with

’known’ d¹³C values (determined with an elemental analyser

coupled to an isotope ratio mass spectrometer; EA/IRMS), so

long as the reference AAs have been derivatised with the

same reagents and analysed concurrently.^[21,27,29]A range

of derivatives are available for AA d¹³C value deter-

minations.^[30–32]Amongst them, we believe that the

N-acetyl methyl ester (NACME) derivative, which adds the

theoretically lowest number of carbon atoms per AA (one

carbon atom to the C-termini and hydroxyl groups, and two

carbon atoms to the N-termini or the hydroxyl moiety), is

the optimal choice in terms of reducing analytical error.^[33,34]

For many years, N-triﬂuoroacetyl isopropyl (TFA/IP) esters

have been widely used for the GC/C/IRMS analysis of

AAs, primarily because of their stability, facile production,

and favourable chromatography.^[30]As a halogen-bearing

derivative, however, TFA/IP has been reported to be detri-

mental to GC/C/IRMS performance,^[26,30,35]but it is still

widely employed as a common derivative for the stable iso-

tope analysis of AAs by GC/C/IRMS.^{[9,10,22,32,36]}

In recent years, liquid chromatography–isotope ratio mass

spectrometry (LC/IRMS) has been introduced as an

alternative method of obtaining AA d¹³C values,^{[8,12,14,37,38]}

and has been applied in a number of areas within life sciences

research.^[39]Derivatisation is not required for LC/IRMS

analysis as the AAs are separated by aqueous HPLC and

oxidised on-line, which reduces the analytical error in d¹³C value

determinations relative to that of GC/C/IRMS.^[40]LC/IRMS

can be used in one of two sample introduction modes: (i)

’HPLC mode’, whereby AAs are resolved and analysed via

IRMS, and (ii) ’bulk mode’, whereby AAs bypass the HPLC

column and enter the IRMS ion source directly, resulting in

a d¹³C value that represents all of the sample carbon.^[12]

Godin et al. used LC/IRMS to separate and analyse

underivatised AAs by a two-dimensional (2D)-LC approach,

coupling a strong cation-exchange column to a reversed-

phase column and using an aqueous buffer throughout.^[41]

This method resulted in the baseline separation of eleven of

ﬁfteen AAs, with a d¹³C value precision better than ꢀ0.2%

at natural abundance, and slightly worse precision for ¹³C-

labelled AAs. The chromatographic separation of AAs for

LC/IRMS analysis has since been improved using mixed-

mode chromatography, combining ion-exchange and

reversed-phase moieties.^[12,37,38]Most published LC/IRMS

methods employ two-phase mobile phase gradients to

separate AAs using mixed-mode chromatography, but a

three-phase system has been shown to improve the chromato-

graphic separation of AAs at the expense of the time required

per sample run.^[14]

In the only methodological comparison of GC/C/IRMS

and LC/IRMS to date, Godin et al. obtained d¹³C values for

valine (Val) in its underivatised form (LC/IRMS) and as N

(O,S)-ethoxycarbonylethyl

ester

derivatives

(GC/C/

IRMS).^[42]The authors compared the techniques over a range

of d¹³C values (ꢁ12.3 to 150.8%) and found comparable root-

mean-square values of standard deviation for GC/C/IRMS

and LC/IRMS. The application of the two techniques to the

analysis of protein-bound Val also showed no statistical

difference (p <0.05) in the isotopic enrichments recorded.

The authors suggest that the addition of derivative carbon

and the necessary correction to the measured d¹³C value

introduced uncertainties for the GC/C/IRMS analysis that

were avoidable through the use of LC/IRMS.

Beyond Godin et al.,^[42]there are no direct comparisons of

GC/C/IRMS and LC/IRMS methodologies for the analysis

of bone collagen across a range of AAs. In this context, this

paper aims to:

i. Develop and validate a method for the determination of

AA d¹³C values by LC/IRMS using standard AAs of

known carbon isotopic composition.

ii. Compare the LC/IRMS method to the most up-to-date

GC/C/IRMS method using the NACME derivative

developed in our laboratory, by determining a range of

AA d¹³C values for a suite of archaeological skeletal collagen

hydrolysates from humans excavated in South Africa.

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

iii. Compare the LC/IRMS- and GC/C/IRMS-derived

AA d¹³C values (prepared as NACME derivatives for

GC/C/IRMS) of the archaeological samples with the pre-

viously published data in Corr et al.,^[9]where collagen

hydrolysates were determined as TFA/IP derivatives by

GC/C/IRMS.

and heated at 110 ^ꢂC for 24 h. Hydrolysates were dried under

nitrogen at 60 ^ꢂC, and re-dissolved in 1 mL of HPLC-grade

water for LC/IRMS analysis. An aliquot of this sample was

dried with subsequent applications of dichloromethane

(DCM) and stored in methanol to await derivatisation as

NACME for GC and GC/C/IRMS analysis.

The interpretations and discussions will focus on a

number of comparisons, including (i) the mass of sample

carbon required for carbon isotopic analysis; (ii) chromato-

graphic performance; (iii) the accuracy and precision of

obtained AA d¹³C values; (iv) the reconstruction of bulk

collagen d¹³C values based on the obtained AA d¹³C values

as a means of evaluating the accuracy of the AA d¹³C values;

(v) the utility of the prevailing Δ¹³C_{Glycine-Phenylalanine}proxy

for the identiﬁcation of marine protein; and (vi) proposed

explanations for observed differences in AA d¹³C values, par-

ticularly for the GC/C/IRMS analysis of TFA/IP derivatives,

which we ﬁnd to be problematic.

AA derivatisation

The NACME AA derivatives were prepared as described in

Corr et al.^[33,34]Brieﬂy, samples were methylated in 1 mL of

an acidiﬁed methanol solution (1.85 M; prepared by the

addition of 800 mL of acetyl chloride to 5 mL of anhydrous

methanol in an ice bath) at 70 ^ꢂC for 1 h. After the removal

of the reagents under a stream of nitrogen in an ice bath,

esteriﬁed AAs were acetylated in a 1 mL solution of acetic

anhydride, triethylamine, and acetone (1:2:5, v/v/v) at

60 ^ꢂC for at least 10 min. The samples were dried under

nitrogen in an ice bath and re-dissolved in 2 mL of ethyl

acetate. A saturated NaCl solution (1 mL) was added and

the organic phase removed upon separation. The samples

were again dried under nitrogen and prepared for GC/C/

IRMS injection.

EXPERIMENTAL

The South African human skeletal collagen hydrolysates

had previously been analysed as N-triﬂuoroacetyl isopropyl

esters.^[9]The AAs were esteriﬁed in acidiﬁed isopropanol

(0.5 mL, 2.8 M with acetyl chloride) and heated at 100 ^ꢂC for

1 h. Excess reagent was evaporated under nitrogen at 40 ^ꢂC,

with subsequent applications of DCM. The esteriﬁed AAs

were dissolved in triﬂuoroacetic anhydride (TFAA, 0.5 mL)

and DCM and heated at 100 ^ꢂC for at least 10 min. Excess

reagent was removed under nitrogen and the AA derivatives

were diluted for GC/C/IRMS injection.

Methods and materials

Standard DL-AAs were purchased from Sigma Aldrich

(Gillingham, UK) and used to prepare equimolar mixtures

and a mixture resembling human type-I collagen. The bulk

d¹³C and d¹⁵N values of the standard AAs were deter-

mined relative to a caffeine standard (d¹³C = ꢁ38.23%;

d¹⁵N = ꢁ18.42%) by EA-IRMS by Thermo Scientiﬁc in

Bremen, Germany. All purchased reagents were of analytical

grade.

Ten archaeological human bone collagen AA isolates were

obtained from a range of sites in South Africa, all of which

had previously been analysed via GC/C/IRMS as TFA/IP

derivatives as part of a larger study.^[9]The ten samples fall

into three broad diet groups on the basis of previous bulk

and compound-speciﬁc stable isotope data, and available

archaeological information: (i) high marine protein (HMP)

consumers (4 individuals); (ii) mixed terrestrial consumers

within a C₄-dominated ecosystem (3 individuals); (iii) mixed

terrestrial consumers within a C₃-dominated ecosystem (3

individuals). Diet group assignments, site locations, and bulk

EA/IRMS analysis

The bulk skeletal collagen d¹³C and d¹⁵N values of the

archaeological humans from South Africa were determined

on a Finnigan-MAT 252 mass spectrometer coupled to a

Carlo-Erba NA1500 elemental analyser (Thermo Fisher

Scientiﬁc, Loughborough, UK).^[43]A variety of standards

were used, including ANU sucrose (d¹³C = ꢁ10.6%; NIST,

Gaithersburg, MD, USA) and Merck gelatine (d¹³C =

ꢁ26.80%; d¹⁵N = +12.14%; Merck, Modderfontein, South

Africa), which were calibrated to VPDB with the certiﬁed

reference materials NBS 18, NBS19, and NBS 20 (IAEA,

Vienna, Austria). The replicate measurement reproduci-

bility was ꢀ0.1% for d¹³C values and ꢀ0.2% for d¹⁵N

values.

d

¹³C and d¹⁵N values for these samples are presented in the

Results and Discussion section. The archaeological skeletal

collagens were analysed for their constituent AA d¹³C values

via LC/IRMS and GC/C/IRMS (as NACME derivatives).

Collagen extraction

The surfaces of archaeological bone samples (1–2 mg) were

manually removed and the cleaned bone demineralised in a

dilute HCl (2%) solution. The bone was subsequently dis-

solved in a 0.1 M NaOH solution overnight and repeatedly

rinsed in distilled water until neutral. The demineralised bone

was freeze-dried and prepared for hydrolysis.^[9]

GC/C/IRMS

The NACME AA d¹³C values were determined using a

Thermo Scientiﬁc Trace GC Ultra gas chromatograph coupled

to a Thermo Scientiﬁc DELTAV Plus isotope ratio monitoring

mass spectrometer via a Thermo GC IsoLink and Conﬂo IV

interface (Thermo Fisher Scientiﬁc). Automated sample

injection was on-column (PAL autosampler; CTC Analytics,

Zwingen, Switzerland). The gas chromatograph was ﬁtted

with a Varian VF23ms capillary column (60 m, 0.32 mm i.d.,

0.15 mm ﬁlm thickness; Varian Ltd., Oxford, UK), and the

following oven temperature programme used: 50 ^ꢂC initial

Collagen hydrolysis

Approximately 2 mg of collagen was placed in a Pyrex culture

tube and dissolved in 6 M HCl. The tubes were ﬂushed with

nitrogen, sealed rapidly with a PTFE-lined screw-top cap,

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

temperatur_ꢁe₁ramped to 120 ^ꢂC at 15 ^ꢂC min^ꢁ1; ramp to 190 ^ꢂC

at 3 ^ꢂC min ; ramp to 250 ^ꢂC at 5 ^ꢂC min^ꢁ1with an isother-

mal hold for 20 min. Helium was used as the carrier gas. Data

were analysed using Isodat NT 3.0 software (Thermo Fisher

Scientiﬁc). All d¹³C values are reported relative to a reference

CO₂of known carbon isotopic composition (d¹³C = ꢁ34.14%),

from a different source from that used for LC/IRMS, which

was introduced directly into the ion source in three pulses

at the beginning and end of each run. Norleucine (Sigma-

Aldrich Co. Ltd.), a non-naturally occurring AA, was added

as an internal standard to each collagen hydrolysate.

ꢁ26.1%; d¹³C_Pro= ꢁ10.6%). Data were analysed with Isodat

software (version 2.5; Thermo Fisher Scientiﬁc). All analyte

peaks were manually integrated to ensure that the d¹³C

values were calculated accurately.

RESULTS AND DISCUSSION

The following comparison of LC/IRMS and GC/C/IRMS for

the analysis of archaeological collagen involves many

considerations, including: (i) the mass of analyte carbon

required for analysis; (ii) the chromatographic performance;

(iii) the accuracy of the AA d¹³C values; (iv) the accuracy of

the reconstructed bulk collagen d¹³C values from indivi-

dual AA d¹³C values; (v) the utility of the established

Δ¹³C_{GlycineꢁPhenylalanine}proxy for marine food consumption;

and (vi) the exploration of interpretive frameworks to explain

any observed differences in d¹³C values produced by different

analytical techniques.

The TFA/IP AA d¹³C values were determined using an HP

6890 gas chromatograph (Hewlett Packard, Santa Clara, CA,

USA) coupled to a ThermoFinnigan MAT DELTA^plusXL

isotope ratio mass spectrometer via a Thermo Finnigan

MAT GC III combustion interface (both from Thermo Fisher

Scientiﬁc), as described in Corr et al.^[9]The amino acids were

separated using a ZB-5 ms column (60 m, 0.32 mm i.d.,

0.25 mm ﬁlm thickness; Phenomenex, Macclesﬁeld, UK), and

the following oven temperature programme employed:

isothermal hold at 40 ^ꢂC for 1 min; r_ꢁa₁mp to 70 ^ꢂC at 10 ^ꢂC

min^ꢁ1; ramp to 170 ^ꢂC at 10 ^ꢂC min ; ramp to 250 ^ꢂC at

Mass of analyte carbon: comparing LC/IRMS and

GC/C/IRMS

15 ^ꢂC min^ꢁ1

.

Correction for additive derivative carbon atoms and

induced kinetic isotope effects (KIEs) for both GC/C/IRMS

methods was carried out using standard techniques

involving mass balance equations and the determination of

AA-speciﬁc correction factors.^[28,29,44]Brieﬂy, standard AAs

of known isotopic value (measured by EA/IRMS) were

derivatised using the same reagents as used for the archaeolo-

gical samples to determine correction factors. Upon obtaining

GC/C/IRMS-derived AA d¹³C values for the derivatised

AAs of the archaeological humans, the correction factors were

used to calculate the d¹³C values of the underivatised AAs.

This process leads to error propagation that is avoided in

the LC/IRMS analysis of underivatised AAs.

Approximately 6 mg of collagen hydrolysate were injected on-

column for the LC/IRMS analyses of archaeological skeletal

collagen. This equates to roughly 2440 ng of total carbon

and means that Gly, the most abundant bone collagen AA

(33% of all AAs), is represented by about 635 ng of carbon.

The least abundant AA, Tyr (0.47% collagen carbon

contribution), only represented 11.5 ng of carbon, assuming

that it was not destroyed during the 6 M HCl acid hydrolysis

(100 ^ꢂC) of the protein. Both extremes fall within the

recommended range for optimal LC/IRMS performance

(8–1320 ng of carbon for each AA), as deﬁned by Smith et al.^[14]

GC/C/IRMS analysis requires AA derivatisation, which

was performed on a sub-sample (400 mL) of the collagen

hydrolysate investigated by LC/IRMS. Upon drying under

nitrogen, the derivatised AAs were dissolved in 1 mL of sol-

vent, diluted 1:10 to a volume of 1 mL, and 1 mL of this solu-

tion was injected into the GC/C/IRMS system. This injection

volume contained approximately 120 ng of total collagen car-

bon. Thus, the GC/C/IRMS methodology requires a much

smaller sample size (roughly 5% of that required for LC/

IRMS analyses), which could be of particular value for small

fossils or very old or poorly preserved bones, where the col-

lagen yields are around 1% skeletal tissue by mass.

LC/IRMS

The collagen hydrolysates were analysed using a Surveyor

liquid chromatograph connected to a Delta V Plus mass

spectrometer via an LC Isolink interface (all from Thermo

Fisher Scientiﬁc).^[40]The AAs were resolved on a Primesep

A column (250 mm ꢃ 3.2 mm, 5 mm particle size, 100 Å pore

size; SiELC Technologies Ltd., Prospect Heights, IL, USA)

using

a

linear two-phase mobile phase gradient

progressing from 100% water to 100% 0.024 M H₂SO₄.

The reference CO₂gas was calibrated against the IAEA

standard L-glutamic acid (USGS40; d¹³C = ꢁ26.24%;

Seibersdorf, Austria). Analyte runs were bracketed after

every four to ﬁve runs with a ﬁve-AA standard mix

consisting of underivatised single AAs purchased from Sigma

Aldrich, which represent a wide range of d¹³C values for

natural abundance determinations (d¹³C_Asp= ꢁ22.4%;

Chromatographic performance

Typical LC/IRMS and GC/C/IRMS chromatograms of the

same archaeological collagen hydrolysates are presented in

Fig. 1. The elution order of the AAs is very different between

the two methods, as a result of the different underlying

separation mechanisms. The GC/C/IRMS separation was

achieved using a 100% cross-bonded cyanopropyl phase

d¹³C_Glu(USGS40) = ꢁ26.24%; d¹³C_Gly= ꢁ32.4%; d¹³C_Ala

=

▶

Figure 1. LC/IRMS chromatogram (A) of individual UCT 1691 determined on the m/z 44 cup with the 45/44 ratio. The baseline

shift at ꢄ1600 s is due to the addition of sulphuric acid in the mobile phase. Four reference gas peaks were used to mitigate

intra-run d¹³C value drift. GC/C/IRMS chromatogram (B) of the NACME derivatives of the AAs from the same individual.

Back-ﬂush on until ꢄ750 s to preserve reactor and ﬁlament life. Two reference gas peaks were used and IS refers to the internal

standard (D-norleucine). Partial GC/C/IRMS chromatogram (C) of the TFA/IP derivatives.

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

(Varian VF-23 ms) with helium as the carrier gas. The LC/IRMS

separation used a mixed-mode HPLC column including both

reversed-phase and ion-exchange moieties (SIELC Primesep

A), with an aqueous phase consisting of water and dilute sul-

phuric acid.

GC/C/IRMS separation is governed by the volatility of

analytes and their interaction with the stationary phase. The

cyanopropyl phase is polar, causing the low molecular

weight, non-polar AAs such as alanine (Ala), Gly and leucine

(Leu) to elute ﬁrst, as these do not interact strongly with the

stationary phase. These are followed by the higher molecular

weight, neutral, and the lower molecular weight polar AAs.

The high molecular weight polar and/or aromatic AAs such

as Tyr and lysine (Lys) elute last of all as these are strongly

retained by polar stationary phases.

The width of the peak for Gly, an AA that is fully resolved

using both techniques, can be used to calculate the theoretical

plate numbers of the HPLC and GC columns. Using Eqn. (1),

where tR is the retention time of Gly, and w its baseline peak

width (assuming a Gaussian distribution), and both tR and w

are measured in seconds, one arrives at the theoretical plate

numbers presented in Table 1.^[45]

respectively, all the mean bulk d¹³C values obtained via LC/

IRMS fall within ꢀ0.5% of the EA/IRMS-derived values, sug-

gesting good agreement between the two techniques (Fig. 2).

Differences in the mean d¹³C values are larger between bulk

and compound-speciﬁc LC/IRMS analyses. Compound-

speciﬁc d¹³C values were obtained for two reference AA

mixtures: (i) a 16 AA mixture resembling human Type-I

collagen; and (ii) a mixture containing ﬁve AAs (DL-Asp,

L-Glu (USGS 40), Gly, DL-Ala, DL-Pro), which represent a

d¹³C value range of ꢁ32.42% (Gly) to ꢁ10.64% (DL-Pro)

and contains the IAEA standard USGS40 L-Glu (d¹³

C = ꢁ26.24%). Mixture (ii) was used as an external standard

to bracket the archaeological samples during analyses.

With the exception of DL-Glu, HPLC analyses of the col-

lagen reference produce lower (i.e. more negative) d¹³C

values than EA/IRMS (Fig. 2). This phenomenon has been

observed elsewhere.^[12,46]The majority of on-column LC/

IRMS d¹³C values are within ꢀ1% of the EA/IRMS value,

which is in line with previously published data for reference

AA mixtures.^[14]The unique behaviour of DL-Glu can be

attributed to the character of its elution. The use of two-phase

linear gradients, whilst straightforward, typically introduces

a chromatographic baseline shift that is connected to the

introduction of the acid phase (0.024 M H₂SO₄); this is a

drawback to using two-phase LC gradients. When using the

18 min to 48 min (0% acid to 100% acid) gradient to resolve

the collagen AA mixture, DL-Glu consistently elutes on the

baseline drop, making it difﬁcult to determine its d¹³C value

due to the complexities of peak integration. This also explains

why the LC/IRMS-determined Ser (serine) d¹³C values are

within ꢀ0.5% of its EA/IRMS-derived value. Ser is typically

affected by the baseline ﬂuctuation,^[12,16]but in this case it

consistently elutes after the baseline has stabilised, thereby

making peak integration straightforward. Critically, HPLC

analyses of the ﬁve reference AAs also showed that the L-

Glu d¹³C values can be accurately determined with a stable

baseline (see Fig. 3).

ꢀ

ꢁ

2

tR

w

N ¼ 16

(1)

It is clear that the GC column has potentially higher

resolving power as reﬂected in part by the much higher plate

count, resulting in far shorter chromatographic runs to

separate all the AAs. The GC/C/IRMS analyses run times

were ca. 60 min while those for the LC/IRMS analyses were

ca. 150 min.

Comparison of LC/IRMS-, GC/C/IRMS-, and EA/IRMS-

derived d¹³C values of standard AAs

The range of d¹³C values obtained was relatively wide for

AAs of low abundance in the collagen mixture (e.g.

methionine (Met), histidine (His), Tyr), reﬂecting the difﬁ-

culty of obtaining accurate d¹³C values for low abundance

analytes in mixtures where the various components are pre-

sent in a wide range of concentrations.

As LC/IRMS can be used for bulk and compound-speciﬁc

stable carbon isotope determinations, it is important to

consider whether there is any instrument-induced fractiona-

tion or loss of precision during either mode of analysis.

Bulk-mode LC/IRMS d¹³C values for reference AAs were

obtained by dissolving each AA in HPLC-grade water and

bypassing the HPLC column upon injection. Up to six injec-

tions were made of each AA per run, and each reference

AA d¹³C value is represented by a minimum of six injections,

or a maximum of 18 injections. With the exception of Lys

and proline (Pro), which show offsets of 0.69% and 0.91%,

Repeated injections of the ﬁve reference AAs (equimolar

concentration) show that the vast majority of LC/IRMS-

derived d¹³C values fall within ꢀ1% of the mean EA/IRMS-

determined value (Fig. 3). The largest mean offset was

observed for Pro (ꢁ0.51%); the mean d¹³C values for the

remaining four AAs are within ꢀ0.3% of their respective

EA/IRMS-determined values, in line with the trend exhibited

in Fig. 2. No signiﬁcant differences were observed between

the bulk d¹³C values (direct injections bypassing the HPLC

column) and the compound-speciﬁc d¹³C values (compounds

passed through the HPLC column) during the analysis of the

international L-glutamic acid standard USGS40 over a wide

period of time, including the period in which the

archaeological samples discussed below were determined.

An average d¹³C value of ꢁ25.7% (standard deviation

(SD) = 0.4%, n = 227) was obtained for direct injection LC/

IRMS, whereas a mean d¹³C value of ꢁ26.2 (SD = 0.6%,

n = 48) was obtained for the LC/IRMS d¹³C value

determinations (two sample t-test assuming unequal

Table 1. Number of theoretical plates for the GC and LC

columns described in the text, using the glycine (Gly) peak

from the chromatograms in Fig. 1. tR refers to retention

time and w refers to baseline peak width, both of which

are reported in seconds

tR

(s)

W

(s)

Column

tR/w

N

LC

GC

1925.7

1100.2

91.3

22.4

21.09

49.12

7100

38600

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

Figure 2. Difference between LC/IRMS-derived d¹³C values for reference AAs

in bulk injection mode and: (i) EA/IRMS-derived d¹³C values; (ii) LC/IRMS-

derived d¹³C values in HPLC mode, when reference AAs are analysed as a

collagen-like (type-I collagen) mixture. Bulk mode LC/IRMS d¹³C values are

represented by a minimum of eight determinations (DL-Hyp) and a maximum

of 24 determinations (DL-Phe, DL-Pro, Gly). Mean LC/IRMS d¹³C values in

HPLC mode are represented by a minimum of six determinations (DL-Glu,

Gly, DL-Thr), and a maximum of 18 determinations (DLVal, DL-Ala, DL-Pro, DL-

Met). Error bars represent the reproducibility of LC/IRMS measurements (ꢀ1s).

variance: p >0.05 when hypothesising a mean difference of

≥0.3%, beyond the standard analytical error for LC/IRMS and

EA/IRMS; t_crit(two-tail) = 2.00; df = 58).

population level for the two analytical techniques, the abso-

lute d¹³C values are different for the same individual with

respect to analytical technique. Differences in the AA d¹³C

values obtained by the different techniques range from 0.1 to

2.8%, with the exception of individual UCT 5217, for which

large d¹³C value differences were observed (2.5 to 5.5%).

There is a systematic offset of about 2–3% in the d¹³C values

for Ala.

Accuracy of AA d¹³C values obtained by LC/IRMS and

GC/C/IRMS

Hyp and Pro show very similar d¹³C values for each indi-

vidual. This is to be expected as Hyp is formed by the in situ

hydroxylation of Pro residues in the Y position in the Gly-

X-Y sequence of collagen.^[47,48]This is a good indicator of

instrumental performance and collagen preservation. Of the

Compound-speciﬁc d¹³C values for the collagen AAs of the

individuals from South Africa, determined by LC/IRMS

and GC/C/IRMS (using NACME derivatives), are presented

in Fig. 4, with associated d¹³C values listed in Tables 2 and 3.

Whilst the pattern of the AA d¹³C values is similar at the

Figure 3. Difference between LC/IRMS-derived and EA/IRMS-derived d¹³C

values. Numbers of determinations for each AA are indicated in parentheses.

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

Figure 4. d¹³C values of the major AAs from the archaeological human skeletal

collagen hydrolysates from South Africa, as determined by LC/IRMS and

GC/C/IRMS (as NACME derivatives). Error bars represent ꢀ1 s. Relevant

diet-colour relationships are as follows: white – C₃consumers; grey – HMP

consumers; black – C₄consumers.

seven AAs that can be compared for the two techniques, Phe

tend to produce comparatively high d¹³C values. The pro-

posed mechanisms explaining these results are discussed

below.

consistently has the lowest d¹³C values while Gly has the

highest d¹³C values.

When the obtained AA d¹³C values are presented

as Bland-Altman plots (Fig. 5), it is clear that the vast

majority of the AA d¹³C values are lower for LC/IRMS

than GC/C/IRMS determinations (analysed as NACME

derivatives). This implies that either proportionally less

¹³CO₂, or proportionally more ¹²CO₂, is entering the

mass spectrometer for each AA for LC/IRMS than for

GC/C/IRMS. The nature of these possibilities is discussed

in detail below.

Reconstruction of bulk collagen d¹³C values

Individual AA d¹³C values obtained by GC/C/IRMS and

LC/IRMS can be used to reconstruct bulk collagen d¹³C

values using a mass balance equation that is anchored to the

carbon contribution of the constituent AAs of skeletal col-

lagen, as deﬁned by genomic data.^[16]Three broad mass bal-

ance comparisons of the LC/IRMS and GC/C/IRMS d¹³C

(as NACME derivatives) values, and the original values pre-

sented in Corr et al.,^[9]can be considered. First, only the seven

AAs common to all three methods are considered: Asp, Hyp,

Glu, Gly, Ala, Pro and Phe. Together, these AAs represent

74.9% of collagen carbon. Secondly, all the AA d¹³C values

available for each technique are considered, representing

95.6–96.2% of collagen carbon for LC/IRMS analyses, 86.3%

collagen carbon for the GC/C/IRMS analysis of NACME

derivatives, and 74.9% collagen carbon for the GC/C/IRMS

analysis of TFA/IP derivatives. Thirdly, the GC/C/IRMS-

derived d¹³C values of both derivatives are compared,

normalised to 100% carbon contribution using the following

formula, and relative to the LC/IRMS-derived values:

When one compares the LC/IRMS-derived d¹³C values

with those obtained by Corr et al.,^[9]where the AAs were

analysed via GC/C/IRMS as TFA/IP derivatives, it is clear

that the vast majority of the AA d¹³C values are also lower

(i.e. more negative) for LC/IRMS determinations (Fig. 5(B)).

For Gly, however, most TFA/IP d¹³C values are lower than

the LC/IRMS-determined values; this is the only AA to show

this trend. Although there is reasonable agreement for Ala

and Pro (ca. ꢀ2%), the values typically differ by more

than 2% between the techniques. There are also notable

observed differences between the Pro and Hyp d¹³C values,

when from a metabolic perspective they should be very

similar.

d¹³C_{Norm:MassBal:}¼ ðΣA_iꢅPÞꢅ1=ΣP_i

(2)

A

comparison of the two GC/C/IRMS methods

(TFA/IP vs. NACME derivatives) shows a similar spa-

tial patterning to the comparison between LC/IRMS-

and GC/C/IRMS-derived AA d¹³C values when the GC/

C/IRMS analyte samples are prepared as NACME deriva-

tives (Fig. 5(C)). However, the differences in the d¹³C values

of the TFA/IP and NACME derivatives are larger. Overall,

with the notable exception of Gly, the TFA/IP esters

where A is the d¹³C value of every AA ’I’, P_iis the percentage

carbon contribution (expressed in decimal form) of every AA

’I’, and Σ P_iis the total carbon contribution of all available hydro-

lysate AAs as a fraction of 100% (expressed in decimal form).

A summary of the three comparisons is presented in Table 4,

and the third comparison speciﬁcally is presented in Fig. 6.

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

Figure 5. Bland-Altman plots comparing d¹³C values for the

AAs from archaeological skeletal collagen hydrolysate AAs

common to all three analytical techniques: (A) compares

LC/IRMS and GC/C/IRMS (NACME derivative) determina-

tions; (B) compares LC/IRMS and GC/C/IRMS (TFA/IP

derivative) determinations; (C) compares GC/C/IRMS ana-

lyses of NACME and TFA/IP derivatives. All GC/C/IRMS

determinations are corrected for additive derivative carbon

and induced kinetic isotope effects.

The major difference between the LC/IRMS- and GC/C/IRMS-

reconstructed d¹³C values is the loss of Arg during GC/C/IRMS

sample preparation (ꢄ8% of collagen carbon), as Arg does not

readily derivatise.^[9,31,32]From the LC/IRMS data, it is clear

that the Arg d¹³C values are low relative to the bulk d¹³C

values. Accordingly, a reconstructed bulk collagen d¹³C value

without Arg should be higher than previously expected.

Indeed, the Δ¹³C_{ReconstructedꢁEA/IRMS}values for the GC/C/

IRMS-derived d¹³C values are higher than expected (Table 4).

From the ﬁrst comparison, it is clear that all three techniques

produce roughly similar reconstructed d¹³C values. The LC/

IRMS-derived mass balance d¹³C values are offset from the

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

EA/IRMS d¹³C values by +3.2 to +5.3% (Avg = +4.1%),

whereas the NACME mass balance d¹³C values differ by

+2.9 to +5.9% (Avg = +4.2%) and the TFA/IP mass balance

d¹³C values differ by +2.6% to +5.5% (Avg = +4.2%).

For the second comparison, however, there are observable

differences between the reconstructed and bulk EA/IRMS

d¹³C values. All the reconstructed bulk d¹³C values, calcu-

lated using the LC/IRMS-derived d¹³C values, are within

ꢀ1% of the EA/IRMS d¹³C values. The NACME-derived

bulk d¹³C reconstructions (representing 86.3% of the collagen

carbon) differ from the EA/IRMS-derived d¹³C values by

between +1.5 and 4.4% (Avg = +2.7%). The TFA/IP-derived

bulk d¹³C reconstructions differ from the EA/IRMS-derived

d¹³C values by the most, but they only represent 74.9% of

the collagen carbon.

When the various NACME- and TFA/IP-derived bulk d¹³C

value reconstructions are normalised to 100% collagen carbon

representation (using Eqn. (2) above), many reconstructed

d¹³C values fall within ꢀ1% of the EA/IRMS-determined

d¹³C values, although differences of up to +3.5% are

observed. Fundamentally, the normalised mass balance

calculations yield accurate bulk d¹³C reconstructions (i.e.

ꢀ1% relative to the EA/IRMS determinations) even when

the obtained AA d¹³C values of one of the analytical techni-

ques are comparatively inaccurate. This suggests that whilst

accurate bulk d¹³C values can be reconstructed, this does

not necessarily mean that accurate bulk d¹³C reconstructions

actually lead to accurate AA d¹³C values. Thus, whilst bulk

reconstructions are a worthwhile analytical check, they are

not sufﬁcient by themselves to ensure the production of

accurate AA d¹³C values. Overall, it is clear that the LC/

IRMS-derived d¹³C values provide the best calculated bulk

d¹³C values

Peak area ratios and collagen preservation

Peak area ratios can be used to assess the preservation of

archaeological collagen by determining whether the AA

proﬁle of a collagen hydrolysate resembles that of modern,

intact skeletal collagen. Mean peak area ratios for the LC/

IRMS and GC/C/IRMS (as NACME) analyses, when

normalised to Gly, are presented in Fig. 7.

For the LC/IRMS-derived peak areas, it is clear that the

vast majority of AAs fall between 90 and 100% of their

expected areas relative to theoretical collagen (as deﬁned by

genomic data for human type-I collagen). However, four

AAs have markedly lower peak area ratios: Arg, Met, Lys,

and Ser. The ﬁrst three cases can probably be attributed to

hydrolytic loss.^[49–52]As previously mentioned, Ser often

elutes on the dropping baseline slope when the acid phase

of a two-phase linear gradient is introduced. This not only

makes it difﬁcult to manually integrate Ser peaks, since it

underestimates the peak area because the baseline, whether

manually or automatically applied, is placed on the

horizontal when the actual baseline of the peak is sloping

downwards. This phenomenon explains the lower than

expected peak area ratio for Ser.

For the AA hydrolysates prepared as NACME derivatives,

some components exhibit a higher than expected peak area

ratio (>110%), while for others the ratio is lower than

expected (<90%) peak area ratio: the former include Ala,

Asp, Leu, Phe, and Pro. This phenomenon can probably be

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

Figure 6. Bar plot depicting differences between reconstructed and EA/IRMS-

determined d¹³C values for a suite of archaeological skeletal collagen hydroly-

sates. Reconstructed d¹³C values were calculated using a mass balance equation

and compound-speciﬁc d¹³C values obtained via three analytical techniques:

(i) LC/IRMS analysis of underivatised AAs; (ii) GC/C/IRMS analysis of

NACME derivatives; (iii) GC/C/IRMS analysis of TFA/IP derivatives. All

GC/C/IRMS-determined d¹³C values were corrected for additive derivative

carbon and induced kinetic isotope effects.

Figure 7. Mean peak area ratios (normalised to Gly) for LC/IRMS and GC/C/IRMS analyses

of the archaeological human skeletal collagen hydrolysates from South Africa. Samples ana-

lysed by GC/C/IRMS were prepared as NACME derivatives and all associated peak area

ratios corrected for the added derivative carbon.

attributed to two factors: (i) the relative volatility of Gly

proportionally more Gly is lost relative to other AAs during

the preparation of NACME derivatives. This would explain

why most/ AAs are found in higher than expected abun-

dances. For Pro and Leu, however, co-elution is also a factor.

Leu partially co-elutes with Ile, and Pro partially co-elutes with

Thr. In both cases, the peak areas of Leu and Pro are probably

larger than they otherwise would be if co-elution did not occur.

Four AAs appear in lower than expected abundances when

prepared as NACME derivatives: Lys, Met, Ile, and Thr. For

Thr and Ile this is because part of their peak area has been

subsumed by the neighbouring peak with which it co-elutes.

NACME derivatives, and (ii) analyte co-elution. Although

Gly is the most abundant AA in skeletal collagen, it is also

the lowest molecular weight AA, containing only two carbon

atoms. Gly thus forms the lightest and most volatile NACME

derivative. NACME derivatives are advantageous for obtaining

AA d¹³C values, particularly because so few exogenous car-

bon atoms are added to each analyte AA. However, several

evaporative steps are required to prepare this derivative, and

analyte loss can occur even for the most experienced chemist.

It is suggested that although the overall loss is small,

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

TFA-IP derivatives are signiﬁcantly different, being some

4% higher than from the other methods (LC/IRMS average

d¹³C_Phe= ꢁ21.1 ꢀ1.9%; GC/C/IRMS NACME d¹³C_Phe

=

ꢁ21.6 ꢀ 2.1%; GC/C/IRMS TFA/IP d¹³C_Phe= ꢁ16.8 ꢀ 2.6%).

The mechanisms proposed to explain these observations are

discussed below.

The intra-individual d¹³C value differences between the two

essential AAs Val and Phe are also useful in differentiating

HMP consumers from those utilising terrestrial-dominated

diets, as previously identiﬁed and explained elsewhere

(Fig. 8).^[16]

Analytical error for LC/IRMS and GC/C/IRMS

The error bars (representing ꢀ1 s) for the LC/IRMS data are

visibly smaller than those for the GC/C/IRMS-derived AA

d¹³C values. For example, the AAs Gly and Phe display

average standard deviations of ꢀ0.2 and ꢀ0.3% for LC/

IRMS determinations, respectively, but they are ꢀ0.6 and

ꢀ0.8% for GC/C/IRMS determinations, respectively. These

standard deviations represent the uncertainty in the reported

d¹³C values, and are expected to be different for the different

analytical techniques. LC/IRMS uncertainty arises purely

from instrumental error (ca. ꢀ0.3%), while GC/C/IRMS

uncertainties propagate, and must include the effects of

adding exogenous derivative carbon atoms to the analyte

AAs, any isotopic fractionation induced by non-quantitative

reactions during derivatisation, as well as instrumental error.

Uncertainties associated with correction for the derivative

carbon atoms includes the error in the standard AA d¹³C

values determined off-line by EA/IRMS and the error in the

determination of the Δ¹³C values of those reference AAs as

derivatives by GC/C/IRMS.^[28,29]These additional sources

of error contribute to the overall uncertainty in the GC/C/

IRMS-determined d¹³C values, leading to a higher impreci-

sion in the GC/C/IRMS d¹³C values.

Figure 8. Bar plot of mean Δ¹³C_{Glycine–Phenylalanine}and

Δ¹³C_{Valine–Phenylalanine}values for three human diet groups

(HMP, terrestrial diet in C₄-dominated ecosystem, terrestrial

diet in C₃-dominated ecosystem) as determined by three

analytical methods: (i) LC/IRMS analysis of underivatised

AAs, (ii) GC/C/IRMS analysis of NACME AA derivatives,

and (iii) the GC/C/IRMS analysis of TFA/IP AA derivatives.

The HMP consuming group consists of four individuals

while the other two groups consist of three individuals.

Error bars represent the standard deviation (ꢀ1 s)

of Δ¹³C_{Glycine–Phenylalanine}and Δ¹³C_{Valine–Phenylalanine}value

determinations.

The notably lower abundances of Met and Lys, however, are

probably caused by a combination of hydrolytic loss and

inefﬁcient derivatisation. Indeed, Arg and His are not

detectable in GC/C/IRMS chromatograms.

Utility of paired AA proxies for dietary reconstruction

LC/IRMS and GC/C/IRMS (analysed as NACME

derivatives) exhibit similar intra-individual differences in the

d¹³C values of Gly and Phe (Δ¹³C_{GlycineꢁPhenylalanine}values)

for the human samples (Fig. 8), with both terrestrial consu-

mer groups having lower values than the HMP consumers,

akin to the ﬁndings of Corr et al.^[9]A two-sample t-test shows

that the differences between the Δ¹³C_{GlycineꢁPhenylalanine}

values obtained by LC/IRMS and GC/C/IRMS are statisti-

cally insigniﬁcant (two-sample t-test assuming unequal var-

iance: p = 0.33; t_crit(two-tail) = 2.1; df= 18). The differentiation

using this proxy is akin to previous applications,^[8,9,16,53]and

thus validates its use as an indicator of marine protein con-

sumption. The mean error associated with LC/IRMS

determinations of Δ¹³C_{GlycineꢁPhenylalanine}values is ꢀ0.4%,

while those obtained by GC/C/IRMS, which include com-

pensation for added derivative carbon and induced KIEs (as

described in the Experimental section), have a larger mean

error of ꢀ1.0%.

Explanation of observed AA d¹³C differences for TFA/IP

derivatives

As discussed above, the d¹³C values for Gly as TFA/IP

derivatives are lower than the values obtained with the other

analytical techniques. With the exception of the TFA/IP Ala

and Pro d¹³C values, which show good agreement with the

LC/IRMS-determined values, the remaining AAs have

higher d¹³C values than are obtained with the other two

techniques. We propose a combination of two phenomena

to explain the isotopic differences observed for TFA/IP

derivatives: (i) inefﬁcient analyte combustion, and (ii)

inherent difference in the intra-molecular ¹³C distribution

between AAs.

The effects of combustion inefﬁciencies can be considered

on two timescales: (i) the long term, within in the lifetime of

a GC/C/IRMS combustion reactor, and (ii) the short term,

within the timeframe of a single GC/C/IRMS run.

HF is generated during the combustion of TFA/IP

derivatives, and this is believed to react irreversibly with

the copper wires and nickel catalyst in the GC/C/IRMS

combustion reactor, forming copper and nickel ﬂourides,

respectively; it has been suggested that this process poisons

the platinum catalyst.^[26,35]Ultimately, these factors probably

reduce the efﬁciency of the CuO/NiO reactor. A decrease in

The same trend of differentiation is exhibited by the

TFA/IP d¹³C values, but the absolute Δ¹³C values obtained

for Gly and Phe are different from those from the other two

analytical methods (Fig. 5). The TFA/IP d¹³C values for Gly

are typically lower than from the other two methods, while

the d¹³C values for Phe are considerably higher than from

the other methods. The Phe d¹³C values obtained from the

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

oxidative efﬁciency leads to incomplete oxidation, particularly

for high molecular weight AAs (e.g. Phe). Under these condi-

tions, bonds containing lighter isotopes (i.e. ¹²C) should be

preferentially oxidised, since bonds containing ¹³C are stron-

ger than those consisting of only ¹²C due to the increased acti-

vation energy required for bond cleavage. Incomplete

combustion should generally result in lower d¹³C values than

the expected or ’true’ d¹³C values. The severity of the effects

of combustion inefﬁciencies should increase with time, with

respect to both long- and short-term effects, and may affect

different individual AAs differently, depending on their

elution order.

It has also been observed that TFA/IP derivatives

damage silica capillaries and t-pieces downstream of the

reactor because of HF production.^[34]Should such damage

result in small leaks, detectable by an increase in the argon

background measured (m/z 40), isotopically lighter CO₂may

be preferentially lost to atmosphere, meaning that propor-

tionally more ¹³CO₂enters the isotope ratio mass spectro-

meter, thereby producing higher than expected d¹³C values.

This phenomenon should affect all AA d¹³C values in the

same way, if not to the same degree. Comparing the data in

Tables 2 and 3 with the data in Corr et al.,^[9]it is clear that

the majority of the AAs exhibit higher d¹³C values, enriched

by some 4%, than in the two sets of measurements made

in this study (LC/IRMS and GC/C/IRMS as NACME

derivatives).

after Ala. Gly is also the most abundant bone collagen AA,

resulting in the largest peak area but, as it only contains two

carbon atoms per molecule, it also bears the largest F:C

ratio. Combined, these factors probably contribute to

incomplete combustion, particularly if the combustion

reactor does not have sufﬁcient recovery time after the

combustion of Ala, which elutes immediately before Gly.

Given the preferential combustion of bonds involving lighter

isotopes during incomplete oxidation, the measured d¹³C_Gly

values should be lower than otherwise expected. Indeed,

this is what is found. It is perhaps also of note, relative to

other AAs previously studied in microorganisms, that

Gly has relatively low intra-molecular d¹³C variability.

Admittedly, few intra-molecular AA d¹³C values exist, but

the work of Abelson and Hoering is quite informative on

this matter.^[18]

Phe, Glu, Hyp, Asp

The Phe d¹³C values for the TFA/IP derivatives are higher

than from the two methods compared in this study, as are

the d¹³C values for Glu, Hyp, and Asp, suggesting that a

different mechanism is operating for these AAs. In addition

to eluting in the second half of the GC/C/IRMS chromato-

gram, after many AAs have passed through the oxidation

reactor, these AAs have higher molecular weigh than the

early eluting AAs. They also possess multiple functional

groups (i.e. more than just a C-terminus and N-terminus),

and thus more derivative carbon is added to each of these

AAs. Phe has an aromatic ring that is particularly

difﬁcult to combust. Furthermore, Phe, Glu, and Asp are

known to have large differences in the d¹³C values between

the carboxyl carbon and the carbon atoms comprising the

remainder of the AAs in various microorganisms.^[18]The

carboxyl carbon d¹³C value for these AAs is consistently

higher than those of the residual carbon by multiple per mil

(18.4 ꢀ 5.6%; n = 14).^[18]If these AAs undergo incomplete

combustion as TFA/IP derivatives, where the C atoms in

the molecule around the C-termini are preferentially oxidised,

the resulting d¹³C values should be higher than expected.

Indeed, the d¹³C values for these AAs are consistently higher

than those from the LC/IRMS analyses of the same collagen

samples.

We will now consider interpretive frameworks for each of

the following three TFA/IP AA groups with respect to inefﬁ-

cient sample combustion and the intra-molecular distribution

of their d¹³C values: (i) Ala and Pro; (ii) Gly; (iii) Phe, Glu,

Hyp, Asp.

Ala and Pro

Referring to Fig. 5(B), it is clear that the TFA/IP Ala and Pro

d¹³C values are comparable with the LC/IRMS-derived d¹³C

values. This agreement may be coincidental. However, Ala

is the ﬁrst AA to elute in TFA/IP chromatograms (Fig. 1),

and whilst Pro elutes mid-way through the run, there is a

gap between it and the earlier elution of a series of medium-

to low-abundance collagen AAs (e.g. Ser, Thr, Val, Leu, Ile).

On the timescale of a single GC/C/IRMS run, Ala has the

best opportunity to undergo complete combustion. In a

similar way, the combustion reactor may have sufﬁcient time

to recover before Pro is combusted, ensuring its complete

oxidation. This phenomenon does not explain why the Glu

d¹³C values are consistently offset from the LC/IRMS-derived

values, since the combustion reactor should have the requisite

recovery time for this AA, but, as discussed below, other fac-

tors are probably involved.

Unfortunately, in the absence of intra-molecular d¹³C

values for Pro and Hyp, it is difﬁcult to know how relevant

the intra-molecular distribution of d¹³C values is to under-

standing the isotopic behaviour of Hyp. Hyp is synthesised

in situ from Pro residues, and Pro is in turn biosynthesised

from Glu. As previously shown, the intra-molecular

variability of Glu d¹³C values is known to be relatively large

in microorganisms. From feeding studies and the radiocarbon

dating of individual AAs, it has been shown that Pro tends to

behave like an essential AA, and is primarily routed to body

tissues from the diet, although it is a non-essential AA.^[10,11,54]

Thus, although empirical evidence is lacking, there is reason

to believe that Pro and Hyp should exhibit a relatively large

difference in their d¹³C values between carboxyl and residual

carbon.

Gly

The Gly d¹³C values are lower for the GC/C/IRMS analysis

of TFA/IP derivatives than for the other two analytical

methods. With respect to mass spectrometry, this means that

either proportionally more ¹²CO₂(m/z 44) or less ¹³CO₂(m/z

45) is being detected by the Faraday cups in the isotope ratio

mass spectrometer. Gly is the second AA to elute in the GC/

C/IRMS chromatograms of the TFA/IP derivatives, shortly

Notwithstanding any intra-molecular d¹³C variation, a TFA

group is added to the additional hydroxyl group of Hyp,

making it a larger TFA/IP AA derivative than that of Pro,

and it could be more difﬁcult to combust completely.

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

Assuming that there is a difference in the d¹³C value of the

carboxyl vs. residual carbon in Hyp, and that the d¹³C value

of the carboxyl carbon is higher than that of the total Hyp

carbon (in line with the vast majority of data presented in

Abelson and Hoering ^[18]), one would expect the obtained

Hyp d¹³C values to be higher than if the entire Hyp TFA/

IP derivative was combusted. This explanation is consis-

tent with the obtained TFA/IP d¹³C values, and it remains

the proposed mechanism by which the GC/C/IRMS-derived

d¹³C values were obtained for this AA. It is interesting to

note that whilst the Pro d¹³C values for the TFA/IP deriva-

tives tend to agree with the LC/IRMS-determined d¹³C

values, the Hyp d¹³C values for the TFA/IP derivatives do

not. As described below, it appears that Pro is combusted

completely while Hyp is not.

CONCLUSIONS

The d¹³C values for reference and archaeological skeletal

collagen AAs have been investigated using two analytical

techniques: the LC/IRMS analysis of underivatised AAs

and the GC/C/IRMS analysis of NACME derivatives. The

archaeological skeletal collagen derives from HMP

consuming individuals, and two groups of individuals

that obtained the vast majority of their diet from terrestrial

sources; while both terrestrial consuming groups had

mixed diets, one group inhabited

a

C₃-dominated

ecosystem and the other group inhabited a C₄-dominated

ecosystem. The d¹³C values obtained by both of these

methods were compared with those from previous GC/C/

IRMS analyses of TFA/IP derivatives from the same

samples.^[9]The central ﬁndings of this study are as follows:

NACME as the AA derivative of choice

i. The GC/C/IRMS analysis of NACME derivatives of collagen

AAs requires much less sample carbon than LC/IRMS analy-

sis (6 mg of hydrolysed collagen per LC/IRMS run, 120 ng of

collagen per GC/C/IRMS run); due to higher column

efﬁciencies and thus higher signal-to-noise ratios; the GC/C/

IRMS analysis time was also much shorter (61 min for GC/

C/IRMS compared with 150 min for LC/IRMS).

Whilst notable differences are observed between the LC/

IRMS-determined d¹³C values of underivatised bone col-

lagen AAs and the GC/C/IRMS-determined d¹³C values of

the derivatised AAs as TFA/IP esters, there is broad

agreement between the LC/IRMS-derived data and the

GC/C/IRMS analysis of the NACME derivative. Most

NACME AA d¹³C values are within ꢀ2% of the LC/IRMS

values. The LC/IRMS analysis of reference AAs, in

corroboration with collagen-like peak area proﬁles and the

production of correct mass balance d¹³C value reconstructions

(ꢀ1% of EA/IRMS d¹³C values) at the compound-speciﬁc

level, strongly suggests that the AA d¹³C values produced

by this technique are accurate. By extension, the AA d¹³C

values are also broadly accurate, although differences of over

4% are occasionally observed.

ii. LC/IRMS analyses allow the determination of Arg d¹³C

values, which is not possible by GC/C/IRMS using

NACME or TFA/IP derivatives, resulting in much more

accurate reconstructed bulk collagen d¹³C values due to

the higher proportion of collagen carbon represented

(≥95.5% of collagen carbon for LC/IRMS as opposed

to 86.3% for GC/C/IRMS analyses of NACME

derivatives).

iii. The analytical errors associated with LC/IRMS AA d¹³C

determinations are lower than those for both GC/C/

IRMS methods, particularly because GC/C/IRMS ana-

lyses require correction for additive derivative carbon

and induced KIEs, which introduces additional

uncertainty.

Whilst LC/IRMS is the optimal method for AA d¹³C value

determinations, particularly because AAs can be analysed in

their underivatised form, LC/IRMS is not as widely available

as GC/C/IRMS, although it is becoming more common. For

researchers who are interested in obtaining AA d¹³C values,

but who do not have access to LC/IRMS, it is recommended

that NACME derivatives are employed for GC/C/IRMS.

iv. The Δ¹³C_{GlycineꢁPhenylalanine}values obtained by LC/IRMS

and GC/C/IRMS, using the new NACME derivative,

were statistically indistinguishable and displayed pat-

terns similar to those previously observed, whereby

HMP consumers have larger d¹³C_{GlycineꢁPhenylalanine}

values than terrestrial consumers. However, the absolute

d¹³C_{GlycineꢁPhenylalanine}values for both techniques were

different from those reported by Corr et al.^[9]The Phe

d¹³C values were some 4% higher, while the Gly d¹³C

values were roughly 4% lower as TFA/IP esters. The

interpretive frameworks described here predict that these

offsets occur because of a combination of inefﬁcient com-

bustion and the effects of inherent intra-molecular d¹³C

value distributions. Fundamentally, the use of TFA/IP

derivatives produces inaccurate AA d¹³C values that are

affected by the varying concentrations and retention times

of the AAs present in analyte runs. Although TFA/IP

derivatives are relatively easy to prepare, and have

historically been used with high frequency, we suggest

that this derivative is problematic for determinations at

natural abundance and should be abandoned for such

purposes. We suggest that NACME derivatives, or other

low carbon-containing derivatives that do not contain

halogens, should be used in preference.

NACME AA derivatives have

a high endogenous to

exogenous carbon ratio, and the analytical error associated

with this derivative is smaller than with all other derivatives,

with perhaps the exception of the ethyl esters,^[55]whose

wider application remains untested. Although our recently

introduced NACME amino acid derivatives are highly

volatile, this study conﬁrms that they yield accurate AA

d¹³C values and should be considered the derivative of choice

for GC/C/IRMS analyses.

Whilst TFA/IP derivatives produce the same isotopic

patterns as the other two methods, they do produce inac-

curate AA d¹³C values. Inaccuracies associated with the

GC/C/IRMS analysis of TFA/IP derivatives probably

stem from combustion inefﬁciencies tied to the presence

of ﬂuorine, which is particularly problematic for high

molecular weight AAs with large intra-molecular d¹³C dif-

ferences. Despite these challenges, TFA/IP derivatives con-

tinue to be used widely. For natural abundance

determinations of d¹³C values, where LC/IRMS is not

available, we believe that this derivative should be aban-

doned in favour of NACME, or other comparably low

molecular weight derivatives that do not contain halogens.

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

P. J. H. Dunn, N. V. Honch and R. P. Evershed

collagen: a new method using RP-HPLC. Archaeometry

2001, 43, 421.

[16] N.V. Honch, J.S.O. McCullagh, R.E.M. Hedges. Variation

in amino acid d¹³C values in archaeological humans and

fauna with different dietary regimes: developing frameworks

of dietary discrimination. Am. J. Phys. Anthropol. 2011,

submitted.

[17] Y.I. Naito, N.V. Honch, Y. Chikaraishi, N. Ohkouchi,

M. Yoneda. Quantitative evaluation of marine protein con-

tribution in ancient diets based on nitrogen isotope ratios of

individual amino acids in bone collagen: an investigation

at the Kitakogane Jomon site. Am. J. Phys. Anthropol. 2010,

143, 31.

Acknowledgements

The authors wish to thank Dr Ian Bull (University of Bristol)

for helpful comments on the manuscript and Professor Judith

Sealy (University of Cape Town) for kindly providing the

archaeological samples. We also thank two anonymous

reviewers for their useful comments. This work was

conducted as part of a responsive mode grant from the UK

Natural Environment Research Council (NERC: NE/

D004535/1). The NERC are also thanked for mass

spectrometry facilities.

[18] P.H. Abelson, T.C. Hoering. Carbon isotope fractionation in

formation of amino acids by photosynthetic organisms.

Proc. Natl. Acad. Sci. USA 1961, 47, 623.

REFERENCES

[19] P. Hare, M. Estep. Carbon and nitrogen isotopic composi-

tion of amino acids in modern and fossil collagens. Carnegie

Institution Washington Yearbook 1983, 82, 410.

[20] P.E. Hare, M.L. Fogel, T.W. Stafford Jr, A.D. Mitchell, T.C.

Hoering. The isotopic composition of carbon and nitrogen

in individual amino acids isolated from modern and fossil

proteins. J. Archaeol. Sci. 1991, 18, 277.

[21] S.A. Macko, M.L. Estep, P.E. Hare, T.C. Hoering. Stable

nitrogen and carbon isotopic composition of individual

amino acids isolated from cultured microorganisms. Carne-

gie Institution Washington Yearbook 1983, 82, 404.

[22] S.A. Macko, M.L. Estep, P.E. Hare, T.C. Hoering. Isotopic

fractionation of nitrogen and carbon in the synthesis of

amino acids by microorganisms. Chem. Geol. 1987, 65, 79.

[23] N. Tuross, M.L. Fogel, P.E. Hare. Variability in the preserva-

tion of the isotopic composition of collagen from fossil bone.

Geochim. Cosmochim. Acta 1988, 52, 929.

[1] J. Sealy, in Handbook of Archaeological Science, (Eds: D.R.

Brothwell, A.M. Pollard), John Wiley, London, 2001, pp.

269–280.

[2] B.S. Chisholm, D.E. Nelson, H.P. Schwarcz. Stable-carbon

isotope ratios as a measure of marine versus terrestrial

protein in ancient diets. Science 1982, 216, 1131.

[3] M.J. DeNiro. Stable isotopy and archaeology. Am. Sci. 1987,

75, 182.

[4] M.J. DeNiro, S. Epstein. Carbon isotopic evidence for

different feeding patterns in two hyrax species occupying

the same habitat. Science 1978, 201, 906.

[5] M.A. Katzenberg, in Biological Anthropology of the Human

Skeleton, (Eds: M.A. Katzenberg, S.R. Saunders), John Wiley,

New York, 2000, pp. 305–327.

[6] H. Tauber. ¹³C evidence for dietary habits of prehistoric man

in Denmark. Nature 1981, 292, 332.

[7] R.E.M. Hedges. Isotopes and Red Herrings: Comments on

Milner et al. and Lidén et al. Antiquity 2004, 78, 34.

[8] K. Choy, C.I. Smith, B.T. Fuller, M.P. Richards. Investigation

[24] A. Barrie, J. Bricout, J. Koziet. Gas chromatography stable

isotope ratio analysis at natural abundance levels. Biomed.

Mass Spectrom. 1984, 11, 583.

[25] M.H. Engel, S.A. Macko, J.A. Silfer. Carbon isotope compo-

sition of individual amino acids in the Murchison meteorite.

Nature 1990, 348, 47.

[26] W. Meier-Augenstein. Applied gas chromatography

coupled to isotope ratio mass spectrometry. J. Chromatogr.

A 1999, 842, 351.

[27] J.A. Silfer, M.H. Engel, S.A. Macko, E.J. Jumeau. Stable

carbon isotopic analysis of amino-acid enantiomers by con-

ventional isotope ratio mass-spectrometry and combined

gas-chromatography isotope ratio mass spectrometry. Anal.

Chem. 1991, 63, 370.

of amino acid

d

¹³C signatures in bone collagen to

reconstruct human palaeodiets using liquid chromatogra-

phy-isotope ratio mass spectrometry. Geochim. Cosmochim.

Acta 2010, 74, 6093.

[9] L.T. Corr, J.C. Sealy, M.C. Horton, R.P. Evershed. A novel

marine indicator utilising compound-speciﬁc bone collagen

amino acid d¹³C values of ancient humans. J. Archaeol. Sci.

2005, 32, 321.

[10] M.R. Howland, L.T. Corr, S.M.M. Young, V. Jones, S. Jim,

N.J. van der Merwe, A.D. Mitchell, R.P. Evershed. Int.

J. Osteoarchaeol. 2003, 13, 54.

[11] S. Jim, V. Jones, S.H. Ambrose, R.P. Evershed. Quantifying

dietary macronutrient sources of carbon for bone collagen

biosynthesis using natural abundance stable isotope analy-

sis. Br. J. Nutr. 2006, 95, 1055.

[12] J.S.O. McCullagh, D. Juchelka, R.E.M. Hedges. Analysis of

amino acid ¹³C abundance from human and faunal bone

collagen using liquid chromatography/isotope ratio mass

spectrometry. Rapid Commun. Mass Spectrom. 2006, 20, 2761.

[13] M. Raghavan, J.S.O. McCullagh, N. Lynnerup, R.E.M.

Hedges. Amino acid d¹³C analysis of hair proteins and bone

collagen using liquid chromatography/isotope ratio mass

spectrometry: paleodietary implications from intra-

individual comparisons. Rapid Commun. Mass Spectrom.

2010, 24, 541.

[28] G. Rieley. Derivatization of organic compounds prior to gas

chromatographic-combustion-isotope ratio mass spec-

trometry analysis: identiﬁcation of isotope fractionation

processes. Analyst 1994, 119, 915.

[29] G. Docherty, V. Jones, R.P. Evershed. Practical and theoreti-

cal considerations in the gas chromatography-combustion-

isotope ratio mass spectrometry d¹³C analysis of small

polyfunctional compounds. Rapid Commun. Mass Spectrom.

2001, 15, 730.

[30] W. Meier-Augenstein, in Handbook of Stable Isotope Analytical

Techniques, vol. 1, (Ed: P.A. de Groot), Elsevier, Oxford, 2004,

pp. 153–181.

[31] C.C. Metges, K.-J. Petzke, U. Henning. Gas chromatogra-

phy/combustion/isotope ratio mass spectrometric

comparison of N-acetyl and N-pivaloyl amino acid esters

to measure ¹⁵N isotopic abundances in physiological sam-

ples: a pilot study on amino acid synthesis in the upper gas-

tro-intestinal tract of minipigs. J. Mass Spectrom. 1996, 31,

367.

[14] C.I. Smith, B.T. Fuller, K. Choy, M.P. Richards. A three-phase

liquid chromatographic method for d¹³C analysis of amino

acids from biological protein hydrolysates using liquid

chromatography-isotope ratio mass spectrometry. Anal. Bio-

chem. 2009, 390, 165.

[15] T.C. O’Connell, R.E.M. Hedges. Isolation and isotopic ana-

lysis of individual amino acids from archaeological bone

[32] D.M. O’Brien, M.L. Fogel, C.L. Boggs. Renewable and non-

renewable resources: amino acid turnover and allocation

wileyonlinelibrary.com/journal/rcm

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

Comparison of LC/IRMS and GC/C/IRMS for palaeodietary research

to reproduction in Lepidoptera. Proc. Natl. Acad. Sci. USA

2002, 99, 4413.

[43] J.C. Sealy. Stale carbon and nitrogen ratios and coastal

diets in the Later Stone Age of South Africa: a comparison

and critical analysis of two data sets. Ancient Biomol. 1997,

1, 131.

[33] L.T. Corr, R. Berstan, R.P. Evershed. Development of

N-acetyl methyl ester derivatives for the determination of

d

¹³C values of amino acids using gas chromatography-

[44] J.A. Silfer, M.H. Engel, S.A. Macko. Kinetic fractionation of

stable carbon and nitrogen isotopes during peptide bond

hydrolysis: Experimental evidence and geochemical impli-

cations. Chem. Geol. 1992, 101, 211.

[45] A.D. McNaught, A. Wilkinson, in IUPAC Compendium of

Chemical Terminology, (2nd edn.), International Union of Pure

and Applied Chemistry, Research Triangle Park, USA, 1997.

[46] N.V. Honch. The palaeodietary implications of amino acid

stable isotope analysis: developments in the application of

compound speciﬁc isotope techniques to archaeological

bone collagen. Unpublished DPhil thesis, Oxford University,

Oxford, 2009.

combustion-isotope ratio mass spectrometry. Anal. Chem.

2007, 79, 9082.

[34] L.T. Corr, R. Berstan, R.P. Evershed. Optimisation of deriva-

tisation procedures for the determination of d¹³C values of

amino acids using gas chromatography-combustion/iso-

tope ratio mass spectrometry. Rapid Commun. Mass Spectrom.

2007, 21, 3759.

[35] W. Meier-Augenstein. Use of gas chromatography-combustion-

isotope ratio mass spectrometry in nutrition and metabolic

research. Curr. Opin. Nutr. Metab. Care. 1999, 2, 465.

[36] M. Fogel, N. Tuross. Extending the limits of paleodietary

studies of humans with compound speciﬁc carbon isotope

analysis of amino acids. J. Archaeol. Sci. 2003, 30, 535.

[37] J.S.O. McCullagh, J. Gaye-Siessegger, U. Focken. Determina-

tion of underivatised amino acid d¹³C by liquid chromato-

graphy/isotope ratio mass spectrometry for nutritional

studies: the effect of dietary non-essential amino acid proﬁle

on the isotopic signature of individual amino acids in ﬁsh.

Rapid Commun. Mass. Spectrom. 2008, 22, 1817.

[47] J. Baum, B. Brodsky. Folding of peptide models of collagen

and misfolding in disease. Curr. Opin. Struct. Biol. 1999,

9, 122.

[48] N.M. Green, D.A. Lowther. Formation of collagen hydroxy-

proline in vitro. Biochem. J. 1959, 71, 55.

[49] Z. Bohak. N^e-(DL-2-Amino-2-carboxyethyl)-L-lysine, a new

amino acid formed on alkaline treatment of proteins. J. Biol.

Chem. 1964, 239, 2878.

[38] J.S.O. McCullagh. Mixed-mode chromatography/isotope

ratio mass spectrometry. Rapid Commun. Mass Spectrom.

2010, 24, 483.

[39] J.P. Godin, L.B. Fay, G. Hopfgartner. Liquid chromatogra-

phy combined with mass spectrometry for ¹³C isotopic ana-

lysis in life science research. Mass Spectrom. Rev. 2007, 26,

751.

[40] M. Krummen, A.W. Hilkert, D. Juchelka, A. Duhr,

H.-J. Schlüter, R. Pesch. A new concept for isotope ratio

monitoring liquid chromatography/mass spectrometry.

Rapid Commun. Mass Spectrom. 2004, 18, 2260.

[41] J.P. Godin, J. Hau, L.B. Fay, G. Hopfgartner. Isotope ratio

monitoring of small molecules and macromolecules by

liquid chromatography coupled to isotope ratio mass spec-

trometry. Rapid Commun. Mass Spectrom. 2005, 19, 2689.

[42] J.P. Godin, D. Breuillé, C. Obled, I. Papet, H. Schierbeek,

G. Hopfgartner, L.B. Fay. Liquid and gas chromatography

coupled to isotope ratio mass spectrometry for the determi-

nation of ¹³Cvaline isotopic ratios in complex biological

samples. J. Mass Spectrom. 2008, 43, 1334.

[50] S. Sentheshanmuganathan, A.A. Hoover. Some aspects of

the destruction of lysine under conditions of acid and enzy-

matic hydrolysis of protein materials containing carbohy-

drates. Biochem. J. 1958, 68, 621.

[51] K. Murray, S. Rasmussen, J. Neustaedter, J. Murray Luck.

The hydrolysis of arginine. J. Biol. Chem. 1965, 240, 705.

[52] M.V. Pickering, P. Newton. Amino acid hydrolysis: old pro-

blems, new solutions. LC-GC 1990, 8, 778.

[53] L.T. Corr, M.P. Richards, C. Grier, A. Mackie, O. Beattie,

R.P. Evershed. Probing dietary change of the Kwädąy Dän

Ts’ìnchį individual, an ancient glacier body from British

Columbia: II. Deconvoluting whole skin and bone collagen

d

¹³C values via carbon isotope analysis of individual amino

acids. J. Archaeol. Sci. 2009, 36, 12.

[54] J.S.O. McCullagh, A. Maron, R.E.M. Hedges. Radiocarbon

dating of individual amino acids from archaeological bone

collagen. Radiocarbon 2010, 52, 620.

[55] Y. Chikaraishi, N. Ohkouchi, in Earth, Life, and Isotopes, (Eds:

N. Ohkouchi, I. Tayasu, K. Koba), Kyoto University Press,

Kyoto, 2010, pp. 355–366.

Rapid Commun. Mass Spectrom. 2011, 25, 2995–3011

wileyonlinelibrary.com/journal/rcm

Article Doi

Comparison of liquid chromatography-isotope ratio mass spectrometry (LC/IRMS) and gas chromatography-combustion-isotope ratio mass spectrometry (GC/C/IRMS) for the determination of collagen amino acid δ13C values for palaeodietary and palaeoecological reconstruction

DOI: 10.1002/rcm.5174

Source and publish data:

Authors:

Article abstract of DOI:10.1002/rcm.5174

Full text of DOI:10.1002/rcm.5174

Products guided by the article

R&D Labs maybe for 117426-05-8

Relevant to this article

Hot Product